arXiv subjects

Tao Zhao

Publications and source records attributed to Tao Zhao.

At least 19 recordsLinked to original sources

arXiv2026

Artificial Intelligence for Subsurface Imaging Understanding: A Decade Review of Challenges, Methods, Benchmarks, and Outlook

Subsurface imaging interpretation bridges observed geophysical data and quantitative geological models, supporting hydrocarbon exploration, CO2 storage assessment, and geohazard monitoring. Over the past decade, machine learning and deep learning have substantially reshaped interpretation workflows. This review synthesizes the 2015-2025 literature across four tasks: structural interpretation, geobody identification, seismic facies analysis, and property estimation, tracing the field's evolution from classical machine learning through deep learning to emerging domain foundation models, and how these tasks couple within a single interpretation system. The task remains fundamentally different from other AI applications, facing ambiguous signals, interpretive non-uniqueness, sparse semantics, unfixed target locations, and scarce reliable annotations. We synthesize three defining challenges: interpretation under complex geological conditions, cross-survey semantic generalization under low information density, and the absence of reliable benchmarks. Addressing them will hinge on integrating human expertise, physical constraints, and geological priors into training and inference, and on treating uncertainty quantification as an intrinsic model output. We outline a forward-looking agenda: unified, jointly modelled interpretation systems with cross-task consistency; priors evolving from physics toward language and multimodal supervision; end-to-end uncertainty propagation; human-AI collaboration and agent-orchestrated workflows; and a more rigorous evaluation science supported by an AI-ready data ecosystem. The review is accompanied by an open benchmark resource (CIG-Bench), covering fault segmentation, relative geologic time estimation, geobody segmentation, and property modeling, with synthetic datasets, pretrained baselines, and quantitative evaluation: https://douyimin.github.io/CIG-bench

physics.geo-ph↗

arXiv2026

OP-GRPO: Efficient Off-Policy GRPO for Flow-Matching Models

Post training via GRPO has demonstrated remarkable effectiveness in improving the generation quality of flow-matching models. However, GRPO suffers from inherently low sample efficiency due to its on-policy training paradigm. To address this limitation, we present OP-GRPO, the first Off-Policy GRPO framework tailored for flow-matching models. First, we actively select high-quality trajectories and adaptively incorporate them into a replay buffer for reuse in subsequent training iterations. Second, to mitigate the distribution shift introduced by off-policy samples, we propose a sequence-level importance sampling correction that preserves the integrity of GRPO's clipping mechanism while ensuring stable policy updates. Third, we theoretically and empirically show that late denoising steps yield ill-conditioned off-policy ratios, and mitigate this by truncating trajectories at late steps. Across image and video generation benchmarks, OP-GRPO achieves comparable or superior performance to Flow-GRPO with only 34.2% of the training steps on average, yielding substantial gains in training efficiency while maintaining generation quality.

cs.CV↗

arXiv2026

Alternating Reinforcement Learning for Rubric-Based Reward Modeling in Non-Verifiable LLM Post-Training

Standard reward models typically predict scalar scores that fail to capture the multifaceted nature of response quality in non-verifiable domains, such as creative writing or open-ended instruction following. To address this limitation, we propose Rubric-ARM, a framework that jointly optimizes a rubric generator and a judge using reinforcement learning from preference feedback. Unlike existing methods that rely on static rubrics or disjoint training pipelines, our approach treats rubric generation as a latent action learned to maximize judgment accuracy. We introduce an alternating optimization strategy to mitigate the non-stationarity of simultaneous updates, providing theoretical analysis that demonstrates how this schedule reduces gradient variance during training. Extensive experiments show that Rubric-ARM achieves state-of-the-art performance among baselines on multiple benchmarks and significantly improves downstream policy alignment in both offline and online reinforcement learning settings.

cs.CL↗

arXiv2025

CleanUpBench: Embodied Sweeping and Grasping Benchmark

Embodied AI benchmarks have advanced navigation, manipulation, and reasoning, but most target complex humanoid agents or large-scale simulations that are far from real-world deployment. In contrast, mobile cleaning robots with dual mode capabilities, such as sweeping and grasping, are rapidly emerging as realistic and commercially viable platforms. However, no benchmark currently exists that systematically evaluates these agents in structured, multi-target cleaning tasks, revealing a critical gap between academic research and real-world applications. We introduce CleanUpBench, a reproducible and extensible benchmark for evaluating embodied agents in realistic indoor cleaning scenarios. Built on NVIDIA Isaac Sim, CleanUpBench simulates a mobile service robot equipped with a sweeping mechanism and a six-degree-of-freedom robotic arm, enabling interaction with heterogeneous objects. The benchmark includes manually designed environments and one procedurally generated layout to assess generalization, along with a comprehensive evaluation suite covering task completion, spatial efficiency, motion quality, and control performance. To support comparative studies, we provide baseline agents based on heuristic strategies and map-based planning. CleanUpBench bridges the gap between low-level skill evaluation and full-scene testing, offering a scalable testbed for grounded, embodied intelligence in everyday settings.

cs.RO↗

arXiv2024

Practical Explicit-time Stabilization of a Proportional Control System

Proportional control can be realized directly through the amplification of analog signals, and it also has the advantage of easy tuning parameters in digital signal control. However, it is difficult for the proportional control to preset the upper bound of settling time. To address this problem, a novel practical explicit-time control method is proposed. In bounded initial condition, this method makes this system error converge to a predefined neighborhood of zero within an explicit time. More specifically, the initial condition set and conditionally stable set are solved by practical explicit-time stabilization theorem. Based on that, a proportional feedback control is founded to achieve practical conditional fixed-time stability.

eess.SY↗

arXiv2023

A Multi-robot Coverage Path Planning Algorithm Based on Improved DARP Algorithm

The research on multi-robot coverage path planning (CPP) has been attracting more and more attention. In order to achieve efficient coverage, this paper proposes an improved DARP coverage algorithm. The improved DARP algorithm based on A* algorithm is used to assign tasks to robots and then combined with STC algorithm based on Up-First algorithm to achieve full coverage of the task area. Compared with the initial DARP algorithm, this algorithm has higher efficiency and higher coverage rate.

cs.RO↗

arXiv2022

Structured Attention Composition for Temporal Action Localization

Temporal action localization aims at localizing action instances from untrimmed videos. Existing works have designed various effective modules to precisely localize action instances based on appearance and motion features. However, by treating these two kinds of features with equal importance, previous works cannot take full advantage of each modality feature, making the learned model still sub-optimal. To tackle this issue, we make an early effort to study temporal action localization from the perspective of multi-modality feature learning, based on the observation that different actions exhibit specific preferences to appearance or motion modality. Specifically, we build a novel structured attention composition module. Unlike conventional attention, the proposed module would not infer frame attention and modality attention independently. Instead, by casting the relationship between the modality attention and the frame attention as an attention assignment process, the structured attention composition module learns to encode the frame-modality structure and uses it to regularize the inferred frame attention and modality attention, respectively, upon the optimal transport theory. The final frame-modality attention is obtained by the composition of the two individual attentions. The proposed structured attention composition module can be deployed as a plug-and-play module into existing action localization frameworks. Extensive experiments on two widely used benchmarks show that the proposed structured attention composition consistently improves four state-of-the-art temporal action localization methods and builds new state-of-the-art performance on THUMOS14. Code is availabel at https://github.com/VividLe/Structured-Attention-Composition.

cs.CV↗

arXiv2021

Background-Click Supervision for Temporal Action Localization

Weakly supervised temporal action localization aims at learning the instance-level action pattern from the video-level labels, where a significant challenge is action-context confusion. To overcome this challenge, one recent work builds an action-click supervision framework. It requires similar annotation costs but can steadily improve the localization performance when compared to the conventional weakly supervised methods. In this paper, by revealing that the performance bottleneck of the existing approaches mainly comes from the background errors, we find that a stronger action localizer can be trained with labels on the background video frames rather than those on the action frames. To this end, we convert the action-click supervision to the background-click supervision and develop a novel method, called BackTAL. Specifically, BackTAL implements two-fold modeling on the background video frames, i.e. the position modeling and the feature modeling. In position modeling, we not only conduct supervised learning on the annotated video frames but also design a score separation module to enlarge the score differences between the potential action frames and backgrounds. In feature modeling, we propose an affinity module to measure frame-specific similarities among neighboring frames and dynamically attend to informative neighbors when calculating temporal convolution. Extensive experiments on three benchmarks are conducted, which demonstrate the high performance of the established BackTAL and the rationality of the proposed background-click supervision. Code is available at https://github.com/VividLe/BackTAL.

cs.CV↗

arXiv2020

Equivalent Classification Mapping for Weakly Supervised Temporal Action Localization

Weakly supervised temporal action localization is a newly emerging yet widely studied topic in recent years. The existing methods can be categorized into two localization-by-classification pipelines, i.e., the pre-classification pipeline and the post-classification pipeline. The pre-classification pipeline first performs classification on each video snippet and then aggregate the snippet-level classification scores to obtain the video-level classification score. In contrast, the post-classification pipeline aggregates the snippet-level features first and then predicts the video-level classification score based on the aggregated feature. Although the classifiers in these two pipelines are used in different ways, the role they play is exactly the same---to classify the given features to identify the corresponding action categories. To this end, an ideal classifier can make both pipelines work. This inspires us to simultaneously learn these two pipelines in a unified framework to obtain an effective classifier. Specifically, in the proposed learning framework, we implement two parallel network streams to model the two localization-by-classification pipelines simultaneously and make the two network streams share the same classifier. This achieves the novel Equivalent Classification Mapping (ECM) mechanism. Moreover, we discover that an ideal classifier may possess two characteristics: 1) The frame-level classification scores obtained from the pre-classification stream and the feature aggregation weights in the post-classification stream should be consistent; 2) The classification results of these two streams should be identical. Based on these two characteristics, we further introduce a weight-transition module and an equivalent training strategy into the proposed learning framework, which assists to thoroughly mine the equivalence mechanism.

cs.CV↗

arXiv2019

Estimating Socioeconomic Status via Temporal-Spatial Mobility Analysis -- A Case Study of Smart Card Data

The notion of socioeconomic status (SES) of a person or family reflects the corresponding entity's social and economic rank in society. Such information may help applications like bank loaning decisions and provide measurable inputs for related studies like social stratification, social welfare and business planning. Traditionally, estimating SES for a large population is performed by national statistical institutes through a large number of household interviews, which is highly expensive and time-consuming. Recently researchers try to estimate SES from data sources like mobile phone call records and online social network platforms, which is much cheaper and faster. Instead of relying on these data about users' cyberspace behaviors, various alternative data sources on real-world users' behavior such as mobility may offer new insights for SES estimation. In this paper, we leverage Smart Card Data (SCD) for public transport systems which records the temporal and spatial mobility behavior of a large population of users. More specifically, we develop S2S, a deep learning based approach for estimating people's SES based on their SCD. Essentially, S2S models two types of SES-related features, namely the temporal-sequential feature and general statistical feature, and leverages deep learning for SES estimation. We evaluate our approach in an actual dataset, Shanghai SCD, which involves millions of users. The proposed model clearly outperforms several state-of-art methods in terms of various evaluation metrics.

cs.CY↗

arXiv2019

Cache-Version Selection and Content Placement for Adaptive Video Streaming in Wireless Edge Networks

Wireless edge networks are promising to provide better video streaming services to mobile users by provisioning computing and storage resources at the edge of wireless network. However, due to the diversity of user interests, user devices, video versions or resolutions, cache sizes, network conditions, etc., it is challenging to decide where to place the video contents, and which cache and video version a mobile user device should select. In this paper, we study the joint optimization of cache-version selection and content placement for adaptive video streaming in wireless edge networks. We propose practical distributed algorithms that operate at each user device and each network cache to maximize the overall network utility. In addition to proving the optimality of our algorithms, we implement our algorithms as well as several baseline algorithms on ndnSIM, an ns-3 based Named Data Networking simulator. Simulation evaluations demonstrate that our algorithms significantly outperform conventional heuristic solutions.

cs.NI↗

arXiv2018

Kid-Net: Convolution Networks for Kidney Vessels Segmentation from CT-Volumes

Semantic image segmentation plays an important role in modeling patient-specific anatomy. We propose a convolution neural network, called Kid-Net, along with a training schema to segment kidney vessels: artery, vein and collecting system. Such segmentation is vital during the surgical planning phase in which medical decisions are made before surgical incision. Our main contribution is developing a training schema that handles unbalanced data, reduces false positives and enables high-resolution segmentation with a limited memory budget. These objectives are attained using dynamic weighting, random sampling and 3D patch segmentation. Manual medical image annotation is both time-consuming and expensive. Kid-Net reduces kidney vessels segmentation time from matter of hours to minutes. It is trained end-to-end using 3D patches from volumetric CT-images. A complete segmentation for a 512x512x512 CT-volume is obtained within a few minutes (1-2 mins) by stitching the output 3D patches together. Feature down-sampling and up-sampling are utilized to achieve higher classification and localization accuracies. Quantitative and qualitative evaluation results on a challenging testing dataset show Kid-Net competence.

cs.CV↗

arXiv2016

Observation of Ultralong Valley Lifetime in WSe2/MoS2 Heterostructures

The valley degree of freedom in two-dimensional (2D) crystals recently emerged as a novel information carrier in addition to spin and charge. The intrinsic valley lifetime in 2D transition metal dichalcoginides (TMD) is expected to be remarkably long due to the unique spin-valley locking behavior, where the inter-valley scattering of electron requires simultaneously a large momentum transfer to the opposite valley and a flip of the electron spin. The experimentally observed valley lifetime in 2D TMDs, however, has been limited to tens of nanoseconds so far. Here we report efficient generation of microsecond-long lived valley polarization in WSe2/MoS2 heterostructures by exploiting the ultrafast charge transfer processes in the heterostructure that efficiently creates resident holes in the WSe2 layer. These valley-polarized holes exhibit near unity valley polarization and ultralong valley lifetime: we observe a valley-polarized hole population lifetime of over 1 us, and a valley depolarization lifetime (i.e. inter-valley scattering lifetime) over 40 us at 10 Kelvin. The near-perfect generation of valley-polarized holes in TMD heterostructures with ultralong valley lifetime, orders of magnitude longer than previous results, opens up new opportunities for novel valleytronics and spintronics applications.

cond-mat.mtrl-sci↗

arXiv2015

Software-Defined Hyper-Cellular Architecture for Green and Elastic Wireless Access

To meet the surging demand of increasing mobile Internet traffic from diverse applications while maintaining moderate energy cost, the radio access network (RAN) of cellular systems needs to take a green path into the future, and the key lies in providing elastic service to dynamic traffic demands. To achieve this, it is time to rethink RAN architectures and expect breakthroughs. In this article, we review the state-of-art literature which aims to renovate RANs from the perspectives of control-traffic decoupled air interface, cloud-based RANs, and software-defined RANs. We then propose a software-defined hyper-cellular architecture (SDHCA) that identifies a feasible way of integrating the above three trends to enable green and elastic wireless access. We further present key enabling technologies to realize SDHCA, including separation of the air interface, green base station operations, and base station functions virtualization, followed by our hardware testbed for SDHCA. Besides, we summarize several future research issues worth investigating.

cs.NI↗

arXiv2015

HyCell: Enabling GREEN Base Station Operations in Software-Defined Radio Access Networks

The radio access networks (RANs) need to support massive and diverse data traffic with limited spectrum and energy. To cope with this challenge, software-defined radio access network (SDRAN) architectures have been proposed to renovate the RANs. However, current researches lack the design and evaluation of network protocols. In this paper, we address this problem by presenting the protocol design and evaluation of hyper-cellular networks (HyCell), an SDRAN framework making base station (BS) operations globally resource-optimized and energy-efficient (GREEN). Specifically, we first propose a separation scheme to realize the decoupled air interface in HyCell. Then we design a BS dispatching protocol which determines and assigns the optimal BS for serving mobile users, and a BS sleeping protocol to improve the network energy efficiency. Finally, we evaluate the proposed design in our HyCell testbed. Our evaluation validates the feasibility of the proposed separation scheme, demonstrates the effectiveness of BS dispatching, and shows great potential in energy saving through BS sleeping control.

cs.NI↗

arXiv2014

Energy-Delay Tradeoffs of Virtual Base Stations With a Computational-Resource-Aware Energy Consumption Model

The next generation (5G) cellular network faces the challenges of efficiency, flexibility, and sustainability to support data traffic in the mobile Internet era. To tackle these challenges, cloud-based cellular architectures have been proposed where virtual base stations (VBSs) play a key role. VBSs bring further energy savings but also demands a new energy consumption model as well as the optimization of computational resources. This paper studies the energy-delay tradeoffs of VBSs with delay tolerant traffic. We propose a computational-resource-aware energy consumption model to capture the total energy consumption of a VBS and reflect the dynamic allocation of computational resources including the number of CPU cores and the CPU speed. Based on the model, we analyze the energy-delay tradeoffs of a VBS considering BS sleeping and state switching cost to minimize the weighted sum of power consumption and average delay. We derive the explicit form of the optimal data transmission rate and find the condition under which the energy optimal rate exists and is unique. Opportunities to reduce the average delay and achieve energy savings simultaneously are observed. We further propose an efficient algorithm to jointly optimize the data rate and the number of CPU cores. Numerical results validate our theoretical analyses and under a typical simulation setting we find more than 60% energy savings can be achieved by VBSs compared with conventional base stations under the EARTH model, which demonstrates the great potential of VBSs in 5G cellular systems.

cs.NI↗

arXiv2014

CONCERT: A Cloud-Based Architecture for Next-generation Cellular Systems

Cellular networks are one of the corner stones of our information-driven society. However, existing cellular systems have been seriously challenged by the explosion of mobile data traffic, the emergence of machine-type communications and the flourish of mobile Internet services. In this article, we propose CONCERT (CONvergence of Cloud and cEllulaR sysTems), a converged edge infrastructure for future cellular communications and mobile computing services. The proposed architecture is constructed based on the concept of control/data (C/D) plane decoupling. The data plane includes heterogeneous physical resources such as radio interfacing equipment, computational resources, and software-defined switches. The control plane jointly coordinates physical resources to present them as virtual resources, over which software-defined services including communications, computing, and management can be deployed in a flexible manner. Moreover, we introduce new designs for physical resources placement and task scheduling, so that CONCERT can overcome the drawbacks of the existing baseband-up centralization approach and better facilitate innovations in next-generation cellular networks. These advantages are demonstrated with application examples on the radio access networks (RANs) with C/D decoupled air interface, delay sensitive machine-type communications, and real-time mobile cloud gaming. We also discuss some fundamental research issues arising with the proposed architecture to illuminate future research directions.

cs.IT↗

arXiv2013

Software Defined Radio Implementation of Signaling Splitting in Hyper-Cellular Network

This paper presents the design and implementation of signaling splitting scheme in hyper-cellular network on a software defined radio platform. Hyper-cellular network is a novel architecture of future mobile communication systems in which signaling and data are decoupled at the air interface to mitigate the signaling overhead and allow energy efficient operation of base stations. On an open source software defined radio platform, OpenBTS, we investigate the feasibility of signaling splitting for GSM protocol and implement a novel system which can prove the proposed concept. Standard GSM handsets can camp on the network with the help of signaling base station, and data base station will be appointed to handle phone calls on demand. Our work initiates the systematic approach to study hyper-cellular concept in real wireless environment with both software and hardware implementations.

cs.NI↗